Intelligent Evaluation Model of English Translation Content Quality Based on Improved Neural Network Algorithm

English translation content estimation is a key work in natural language processing. Unlike the conventional automatic evaluation method of English translation content, the translation quality estimation method does not use manual reference translation to evaluate the ability of English translation. However, according to the content quality estimation of the current sentences in English translation, the feature information extraction method lacks the generalization analysis of linguistic research, which also affects the use of subsequent vector regression methods. Therefore, the feature information of the vocabulary vector is studied to obtain the context vocabulary prediction model and matrix analysis model of deep learning. They are combined with the recursive neural network language modeling to enhance the reliability of the independent estimation and manual evaluation of translation quality. The experimental results using the data set of the sub-task of translation content quality estimation in WMT 15 and WMT 16 show that the method of obtaining the feature of sentence vector through context lexical analysis is consistently more effective than the original QuEst method and the feature acquisition method of sentence vector graph in continuous space language mode. It is also clarified that the newly established feature extraction method does not require linguistic means but significantly enhances the effectiveness of translation quality evaluation.